Enhanced thermal design of microchannel heat sinks with oak-shaped pin fins using machine learning and evolutionary algorithms

As microprocessors become more compact and powerful, they generate increasingly high heat fluxes that require effective thermal management. Conventional cooling methods often fall short, prompting the need for advanced solutions. This study presents a numerical investigation of fluid flow and conjug...

Celý popis

Uložené v:

Podrobná bibliografia
Vydané v:	International communications in heat and mass transfer Ročník 169; s. 109595
Hlavní autori:	Fazli, Mohammad Javad, Seraj, Mohammad, Ashjaee, Mehdi, Houshfar, Ehsan
Médium:	Journal Article
Jazyk:	English
Vydavateľské údaje:	Elsevier Ltd 01.12.2025
Predmet:	Artificial neural network CFD Conjugated heat transfer Finned heat sink Multi-objective optimization CFD Multi-objective optimization Artificial neural network Finned heat sink Conjugated heat transfer
ISSN:	0735-1933
On-line prístup:	Získať plný text
Tagy:	Pridať tag Žiadne tagy, Buďte prvý, kto otaguje tento záznam!

Popis
Shrnutí:	As microprocessors become more compact and powerful, they generate increasingly high heat fluxes that require effective thermal management. Conventional cooling methods often fall short, prompting the need for advanced solutions. This study presents a numerical investigation of fluid flow and conjugate heat transfer in a microchannel heat sink equipped with oak-shaped micro pin-fins. The effects of pin-fin diameter (40–75 μm) and inlet Reynolds number (100–900) were analyzed using 3D simulations. Water was used as the working fluid, and a uniform heat flux of 1 MW/m2 was applied to the substrate's bottom surface, modeled as silicon. Thermal performance was characterized by the Nusselt number, while hydrodynamic behavior was assessed via pressure drop. Both metrics were normalized for multi-objective optimization. Two multi-layer perceptron (MLP) neural networks were trained to predict these performance indicators and integrated with the NSGA-II genetic algorithm. This approach enabled efficient identification of trade-offs between heat transfer enhancement and flow resistance. The optimization yielded a Pareto front of solutions, with the best-performing design featuring a 74.25 μm pin-fin diameter and a Reynolds number of 465.8. This configuration demonstrated significantly improved thermohydraulic performance, making it a promising candidate for next-generation microelectronic cooling applications. •Developed an MLP neural network to predict ∆P and Nu in microchannel heat sinks.•Employed NSGA-II for multi-objective optimization of oak-shaped pin-fin geometries.•Demonstrated superior performance with a 72 % increase in Nu at optimal conditions.•Achieved thermohydraulic efficiency improvement by 57 % using optimized pin-fins.•Identified Pareto-optimal solutions balancing heat transfer and pressure drop.
ISSN:	0735-1933
DOI:	10.1016/j.icheatmasstransfer.2025.109595